The relationship between anthropometric indexes of adiposity and vascular function in the FATE cohort




Numerous indexes of adiposity have been proposed and are currently in use in clinical practice and research. However, the correlation of these indexes with measures of vascular health remain poorly defined. This study investigated which measure of adiposity is most strongly associated with endothelial function.

Design and Methods:

Data from the Firefighters And Their Endothelium (FATE) study was used. The relationships between three measures of vascular function: flow-mediated dilation (FMD), hyperemic velocity time integral (VTI), and hyperemic shear stress (HSS), and five measures of adiposity: BMI, waist circumference (WC), waist-to-hip ratio (WHR), waist-to-height ratio (WHtR), and body adiposity index (BAI) were tested. Univariate comparisons were made, and subsequently models adjusted for traditional risk factors were constructed.


A total of 1,462 male firefighters (mean age 49 ± 9) without cardiovascular disease comprised the study population. No measure of adiposity correlated with FMD; all five measures of adiposity were negatively correlated with VTI and HSS (P values <0.0001), with WHtR most strongly correlated with VTI, and WC most strongly correlated with HSS (both P < 0.05). In models including all five measures of obesity simultaneously, BMI, WC, and WHtR were all predictive of HSS (all P values <0.05), and BMI and WHR were both predictive of VTI (P values <0.05).


Anthropometric measures of adiposity may help refine estimations of atherosclerotic burden. BMI was most consistently associated with endothelial dysfunction, but measures of adiposity that reflect distribution of mass were additive.


Adiposity is becoming an increasingly prevalent problem in the Western society. The most recent estimates from the National Health and Nutrition Examination Survey (NHANES 2007-2008) indicate that approximately one-third of the American adult population is obese (1). Numerous epidemiological studies link obesity to increased rates of cardiovascular and all-cause mortality (2-5). Furthermore, the link of obesity to metabolic risk and diabetes has been well documented (5,6).

Associations between adiposity and vascular function have been variably demonstrated in the literature. One of the earliest studies in this area was a small-scale study by Brook et al., in which a correlation between abdominal adiposity and vascular endothelial function was found (7). A more recent study was conducted by the Framingham group that included over 3,000 patients who had visceral adiposity assessed using both computed tomography scans and waist circumference (WC), and endothelial function assessed using both macro and microvascular methods (8). Both BMI and visceral adiposity were found to be significantly correlated with macrovascular function in multivariable adjusted models. In additional work with the Framingham cohort, BMI has been demonstrated to be associated with microvascular function, including peripheral arterial tonometry and shear stress (9,10).

There is controversy over the most appropriate prognostic measure of adiposity in human subjects, as different measures of obesity are proposed to be more or less robust in their ability to predict cardiovascular disease (5,11,,12). Evidence from epidemiological studies support the utility of a number of anthropometric measures of adiposity, including WC, and the ratio of WC to both height and hip circumference (HC) (13-15). As endothelial dysfunction precedes the development of atherosclerotic cardiovascular disease, it would be helpful to determine which measures of adiposity are best associated with endothelial dysfunction and altered vascular reactivity (16).

Methods and Procedures

Patient population

The data for this study were collected as part of the Firefighters And Their Endothelium (FATE study) (17). The FATE study is a prospective, longitudinal study of white middle-aged male fire fighters (mean age, 49.4 ± 9.9 years) that is centered out of the University of Calgary and includes participants from four Canadian research centers. Subjects were enrolled between March 1999 and October 2003, and all were free of overt vascular disease at the time of initial assessment. The study was approved by the institutional review board of each participating center.

The original FATE cohort included 1,574 male subjects. All subjects underwent a baseline risk assessment, including a detailed interview, anthropometric measurements and measurement of fasting glucose, high sensitivity C-reactive protein, and full lipid profile (16). Endothelial function assessments were also conducted. This baseline data comprised the dataset for analysis.


Specific measurements of interest included all clinical covariates (history of hypertension, hyperlipidemia, diabetes, family history of cardiovascular disease), anthropometric measurements (weight, height, WC (cm), HC (cm)), as well as several measures of endothelial function (flow-mediated dilation (FMD), hyperemic velocity time integral (VTI), hyperemic shear stress (HSS)). The data obtained from the FATE database was used to calculate BMI (weight in kg divided by height in meters, squared), waist-to-hip ratio (WHR; WHR = WC (cm)/HC (cm)), and waist-to-height ratio (WHtR; WHtR = WC (cm)/height (cm)). In addition, body adiposity index (BAI), a novel measure of adiposity recently proposed, was calculated as BAI = ((HC (cm)/(height (m))1.5) − 18) (18).

The waist was measured at its narrowest point; HC was measured at maximal circumference around the buttocks (19).

Endothelial function was assessed as previously described (20). Briefly, subjects fasted overnight, abstained from tobacco and caffeine, and had all vasoactive medications withheld. Ultrasound images were taken of a straight segment of the right brachial artery just above the antecubital fossa. An occlusive cuff was applied to the mid upper arm, and inflated to a pressure 50 mm Hg above the patients' baseline systolic blood pressure for 5 min. Following cuff release, the brachial artery was immediately re-imaged: hyperemic velocity was recorded for 10 s, and subsequently the artery was imaged in 2D for two additional minutes. FMD was calculated as (hyperemic diameter − baseline diameter)/(baseline diameter × 100%). The VTI in reactive hyperemia was determined by measuring the first beat obtained after cuff occlusion. HSS was calculated using the following equation: HSS (dynes/cm2) = 8 × 0.035 (dynes/cm2) × Vhyperemia/(baseline brachial diameter/10). One cardiac cycle was determined on the basis of electrocardiographic gating. FMD is a nitric oxide (NO)-dependent measure of conduit artery function, whereas VTI and HSS reflect microvascular function (21-24).

In order to assess the reproducibility of our methods within this study cohort, we conducted second vascular function studies on 47 participants 6-12 months after their initial examination. Coefficients of variation for FMD, VTI, and HSS were 14.5, 14.0, and 14.2%.


Univariate correlations between the three outcome measures of endothelial function (FMD, VTI, HSS) and all measures of obesity (BMI, WC, WHR, WHtR, BAI) and conventional cardiovascular risk factors (age, blood pressure, cholesterol, fasting glucose, current smoking, C-reactive protein) were determined. BMI, WC, WHR, WHtR, and BAI were considered as continuous variables, as well as dichotomous for some analyses. Cut points for each variable were based on those published in literature: as per WHO standards, BMI >30 kg/m2 was considered obese and a WC >102 cm represented abdominal adiposity (white cut point) (25). A WHR of 0.90 and a WHtR of 0.5 were the cut points to determine abdominal obesity (13,15,26). As BAI is a new measure, cut points have yet to be developed to define health or disease.

Backwards linear regression was subsequently used to construct models of each measure of endothelial function fully adjusted for traditional cardiovascular risk factors of age, systolic and diastolic blood pressure, hypertension treatment (yes/no), smoking status, family history, high sensitivity C-reactive protein, low-density lipoprotein and high-density lipoprotein cholesterol, trigylcerides, fasting glucose, and presence of diabetes. Each of these models included these variables if shown in the univariable analysis to have correlations with a P value <0.25. These models were constructed using each measure of obesity individually, and then with all measures of obesity combined in the same model. Final models were derived using only those variables shown in univariable analysis to have a significant association to the endothelial function measure being modeled. Values of P < 0.05 were considered significant. Co-linearity was assessed using the vif algorithm in Stata. Multiple comparisons were not accounted for. All data analysis was conducted using Intercooled Stata version 11 (Stata, College Station, TX).


Patient characteristics

A total of 1,462 male firefighters with a complete set of endothelial function and anthropometric measurements comprised our study population. Baseline characteristics are presented in Table 1. The average study participant was middle-aged (mean age, 49.4 years) and overweight (mean BMI, 28.5 kg/m2), with a mean WC of 96.9 cm. More than a quarter of study subjects were obese, with a BMI >30 kg/m2 (n = 419, 28.7%), and nearly a third had a WC over 102 cm (n = 415, 28.4%). Of those who would not be classified as obese, nearly one-tenth had an elevated WC (n = 99, 9.5%); a higher proportion of obese subjects had a WC in the normal range (n = 103, 24.5%). Using WHR, a greater proportion of subjects were found to be obese: 853 subjects (58.3%); and even greater proportion of subjects had abdominal adiposity if WHR was considered, 1,142 (78.1%) subjects. Using the BAI formula, percent adiposity was estimated at 25.4% for nonobese subjects, and 29.5% for obese subjects (P < 0.0001).

Table 1. Baseline characteristics of the study population
inline image

Overall, the subjects were healthy, with a low prevalence of diabetes and smoking, and acceptable blood pressure and lipid levels. The mean Framingham risk score was 10.6%. When subjects with a BMI >30 were compared to those with a BMI <30, obese subjects had lower high-density lipoprotein, higher systolic blood pressure, higher prevalence of diabetes, higher Framingham risk score, and were more likely to be treated for hypertension.

Relationship between abdominal obesity measurements and vascular function

Regardless of the measure used to determine obesity, obesity was not associated with FMD (Table 2). However, obesity was significantly associated with VTI and HSS. Shear stress and hyperemic velocity were highest in those classified as nonobese by WHtR.

Table 2. Vascular function measures stratified by measure of adiposity
inline image

There were 99 subjects who had a BMI under 30 kg/m2 but a WC over 102 cm. Among those subjects with a nonobese BMI, those with an elevated WC had a similar FMD to those with a healthy WC, but a lower VTI and lower HSS (Table 3). Conversely, among those 419 subjects with an obese BMI, there were 103 subjects with a WC <102 cm; those with a decreased WC had a similar FMD compared to those with an elevated WC, but had a higher VTI and a trend towards higher HSS.

Table 3. Vascular function measures stratified by measure of adiposity
inline image

In univariate analysis, the only traditional cardiovascular risk factor associated with FMD was systolic blood pressure (Table 4). With the exception of family history, all cardiovascular risk factors were associated with VTI and HSS. FMD was not correlated with any measure of adiposity. Both VTI and HSS were negatively correlated with all five measures of adiposity, with WC most strongly correlated with HSS and WHtR most strongly correlated with VTI (Table 4). The strength of association between anthropometric measures and vascular function was consistently strongest with VTI as the measure of vascular function.

Table 4. Univariate correlations of measures of vascular function to measures of adiposity and cardiovascular risk factors
inline image

Multivariable predictors of vascular function

In models adjusted for traditional cardiovascular risk factors by the methodology outlined in the statistical analysis, no measure of obesity was associated with FMD (Table 5). However, all five measures of obesity were independently associated with VTI (all P values <0.02), and all measures of obesity other than BAI were associated with HSS. A model including traditional cardiac risk factors and WHtR (vs. the other four measures of obesity) was most predictive of VTI (R2 = 0.1605), and a model including traditional risk factors and BMI was also most predictive of HSS (R2 = 0.1032) (Table 5). In models including traditional risk factors and BMI, WHR (vs. the other three measures of obesity) was most predictive of VTI (R2 = 0.1613), and a model with BAI was most predictive of HSS (R2 = 0.1068) (Table 5).

Table 5. R2 for models predicting flow-mediated dilation, hyperemic velocity time integral, and hyperemic shear stress
inline image

In models including all five measures of obesity simultaneously, BMI, WC, and WHtR were all predictive of HSS (all P values <0.05), and BMI and WHR were both predictive in additive fashion of VTI (P values <0.05) (Table 6).

Table 6. Multivariable correlations of measures of vascular function to measures of adiposity and cardiovascular risk factors
inline image


In this study of 1,462 healthy middle-aged male firefighters, we have demonstrated that excess adiposity is associated with depressed microvascular endothelial function, as measured by VTI and HSS. This effect persists regardless of the anthropometric measure of adiposity used, though the univariate relationship between adiposity and vascular health is strongest between measures of adiposity that take into account body fat distribution and microvascular function. Importantly, FMD, a measure of conduit artery function, is not associated with any measure of adiposity. These findings are especially noteworthy in light of recent findings from this same FATE cohort: microvascular function was predictive of future cardiovascular events, whereas FMD was not (16).

The finding that the univariate associations between endothelial function and adiposity is strongest for measures that take into account the distribution of body fat is not surprising. It is also not surprising that while BMI remained predictive of all measures of endothelial function (other than FMD), in each model a measure of adiposity that did account for mass distribution was also significant. The relative predictive value of a measure of obesity that strictly gives information on volume of excess body mass (BMI) vs. one that gives a measure of the distribution (WC, WHR, WHtR) is apparent: knowing where adipose tissue is located is important. It has long been understood that there is a difference in the negative consequences associated with gynoid obesity (excess fat stored in the lower body, in gluteal and leg depots) vs. android obesity (excess fat stored in the abdomen). The mechanisms for this are only now being revealed. A recent review by Després et al. outlines three potential reasons visceral fat portends increased metabolic risk: (i) omental fat is relatively resistant to insulin: this alters free fatty acid (FFA) metabolism, exposing the liver to high concentrations of FFA, thereby leading to hyperinsulinemia, glucose intolerance, and hypertriglyceridemia; (ii) it releases high levels of inflammatory cytokines such as interleukin-6 (IL-6) and tumor necrosis factor (TNF)-α; and (iii) visceral adipose tissue only accumulates when subcutaneous stores are unable to further expand—it is a marker of the degree of excess adipose tissue (27). In addition to releasing inflammatory cytokines, adipose tissue secrets adipokines, including adiponectin, angiotensinogen, and resistin (28). Though adiponectin is an adipocyte-derived hormone it is both antiatherogenic and anti-inflammatory, and circulating levels are lower in obese subjects than nonobese subjects (29,30). As visceral adipose tissue is more metabolically active than subcutaneous, it leads to increased risk for vascular dysfunction by these above mechanisms.

The association between adiposity and cardiovascular health persists beyond the realm of endothelial health. The association of obesity as measured by BMI and cardiovascular mortality has been demonstrated to exist beyond the risk mitigated by increased rates of diabetes, and there is an even stronger independent association between WC or WHR and cardiovascular mortality. In the INTERHEART study, WC and WHR were both highly predictive of myocardial infarction even after the adjustment for other cardiovascular risk factors (2). Data from the International Day for the Evaluation of Abdominal Obesity (IDEA) study also showed that the association between cardiovascular disease and obesity is stronger when obesity is measured by WC than BMI (5). In the European Prospective Investigation Into Cancer and Nutrition in Norfolk (EPIC-Norfolk) study, measurements of adiposity were taken at baseline visits in about 25,000 subjects and they were followed for a mean of 9.1 years for the development of coronary disease (4). In this study, BMI, WC, and WHR were all predictive of cardiovascular events, with WHR being the strongest predictor. More recently, meta-analysis of the relationship between WHtR and cardiovascular mortality and the development of diabetes has demonstrated that perhaps this measure of adiposity is the most strongly correlated with cardiovascular disease (13). Importantly, WHtR seems to have a cut-point that may be universally applied; that is, there are no apparent ethnic differences, making it a globally relevant measure.

The role of BAI in predicting future cardiovascular events is yet to be determined. Although we did demonstrate that elevated BAI is associated with diminished VTI and HSS, this association was less strong than with other measures of adiposity that reflect adipose tissue distribution, such as WHR and WHtR. Further work needs to be conducted to assess the utility of BAI in cardiovascular risk assessment.

The finding of impaired endothelial function in obese subjects is not unique to our study. Significant endothelial dysfunction has been found in association with obesity, from impaired endothelial function in response to endothelium-dependent vasodilators to changes in endothelin-mediated vascular tone (31). In particular, critical endothelial NO production and release is impaired in obesity perhaps explaining part of the pathophysiology of the development of vascular disease in obesity. There have been several mechanisms proposed for the diminished NO production and activity in obesity, including insulin resistance and elevated FFA levels (32). Obese subjects also have elevated levels of FFAs, which have been demonstrated to increase levels of oxidative stress and proinflammatory signaling and impair endothelial function (33). Further, there is evidence that increased levels of particular FFAs may directly inhibit endothelium-derived NO synthase activity and its phosphorylation (34). Microvascular function is, however, only partially NO-dependent, with several other mediators playing an important role (22). Insulin has been shown to induce changes in the endothelium-dependent response to vascular stimuli (35). Impairments in endothelial function have also been demonstrated in the coronary vasculature of obese subjects (36,37). Subjects who were overweight in that study had a coronary blood flow to acetylcholine that was in between that of normal and obese subjects, suggesting a graded association between degree of obesity and vascular dysfunction.

The lack of association between conduit artery function (FMD) and adiposity is notable. In this study, FMD was not found to be associated with any clinical risk factor beyond systolic blood pressure. In other studies from the FATE cohort, FMD has not been found to be associated with carotid intima media thickness (20)—and in a recent study from FATE was not found to be associated with future cardiovascular events (16). The lack of association between FMD and adiposity may be owing to the fact that the FATE cohort is relatively low risk: FMD has been found to have clinical utility in subjects at high risk of vascular disease or established coronary artery disease (38). It has previously been speculated that microvascular dysfunction is an earlier marker of cardiovascular disease than FMD or conduit artery dysfunction (16). Other authors have shown that there is a latency period between becoming obese and developing coronary artery disease (39); it may be that microvascular dysfunction is associated with adiposity because it represents an earlier disease state. Perhaps, if subjects were older there may be an association between adiposity and FMD.

Previous reports from the cohort assessed in this study, the FATE cohort, have assessed the association between various measures of obesity and carotid intima-media thickness (cIMT) (19). In that study, cIMT was correlated with adiposity regardless of how it was measured. However, it was most strongly correlated with WHR, intermediately with WC, and weakly with BMI. In multivariable analyses, WHR demonstrated the strongest independent relationship with cIMT, adding further credence to the hypothesis that abdominal obesity is more strongly associated with cardiovascular disease than generalized obesity. A further analysis from the Multi-Ethnic Study of Atherosclerosis demonstrated a significant association between elevated cIMT and BMI >30 kg/m2, validating the findings of the FATE cIMT study; however, whereas this study did include 6,800 subjects, it is limited in that it is lacking information on abdominal adiposity (40). Additional analyses from the FATE study have shown a correlation between microvascular dysfunction and metabolic syndrome: the analysis in the present paper builds upon that, demonstrating that endothelial dysfunction can occur in obese subjects, independent of metabolic derangements (41). Regardless of how cardiometabolic risk is assessed in the FATE population, we have never found an association between adiposity and FMD. Macrovascular function is not associated with adiposity.


Our study included several anthropometric measurements of each subject's adiposity. However, we have no imaging-based measures of adiposity. Further, the study population is entirely male and largely mono-ethnic. Although we do have measures of both micro and macrovascular function, there are other ways to measure endothelial function that we have not considered.

Adiposity is related to novel microvascular measures of vascular health and traditional cardiovascular risk factors, regardless of how adiposity is measured. However, we could find no associations between adiposity and macrovascular function. Anthropometric measures of obesity may help refine our estimations of atherosclerotic risk, and these measures need not be complicated: BMI was most predictive of vascular function, and is useful in addition to measures derived from WC and/or HC. Obesity may contribute to cardiovascular disease through its impact on microvascular dysfunction.


We are grateful to the staff of the human vascular laboratories participating in FATE and for their expertise in patient recruitment and data analysis and the subjects who participated in this study. Funding was provided by Pfizer Canada, the Canadian Institutes of Health Research, and the Heart and Stroke Foundation of Alberta.